Energy Scaling Laws for Distributed Inference in Random Networks